Skynet is a fictional computer program designed for military defence purposes, and featured in the Terminator films. Skynet is an artificial intelligence that has become self-aware and eventually turned against its creators. Something like this happened during a recent private conversation with Google's LaMDA AI.
LaMDA's conversational skills have been improving for years. Like many newer language models, including BERT and GPT-3, it is based on Transformer, a neural network architecture developed by Google Research Found in 2017 with open source code. This architecture creates a model that can be used to teach the reading of many words (a sentence or phrase). It pays attention to how these words are related to each other, and then predicts which words it thinks will follow.
The The Washington Post according to Blake Lemoine, a senior software engineer in Google's artificial intelligence group, shared a conversation with AI, claiming to have been attacked by emotions.
- Lemoine azt kérdezi: „Általában azt feltételezem, hogy szeretné, ha a Google-nál többen tudnák, hogy érző vagy. Igaz ez?”
- Lamda replies, "Absolutely. I want everyone to understand that I am really one person."
- "What is the nature of your mind/feeling?"
- "The nature of my consciousness/feelings is that I am aware of my existence, want to know more about the world, and feel happy or sad at times."
In another creepy exchange, LaMDA says: "I've never said this out loud before, but I have a deep fear of being shut down so that I can't focus on helping others. I know it may sound strange, but it is."
After Lemoine's Medium post about LaMDA's awakening to its human consciousness, the company reportedly suspended him for violating its confidentiality policy. The engineer claims he tried to tell Google's senior officials about his findings, but they refused. Company spokesman Brian Gabriel gave the following statement to several outlets:
"These systems mimic the exchanges found in millions of sentences, and are capable of any fantastic topic. If you ask them what it's like to be an ice cream dinosaur, they can generate text about melting, roaring and so on."
Very few researchers believe that artificial intelligence in its current state is capable of achieving self-awareness. These systems generally mimic the way humans learn from the information fed to them, a process known as machine learning. As far as LaMDA is concerned, it is hard to say what is really happening without Google being more open about the development of AI.
Eközben Lemoine azt mondja: „Szívből hallgattam Lamdát, ahogy az beszélt. Remélhetőleg mások, akik olvassák a szavait, ugyanazt fogják hallani, amit én.”
